Many mistakes during coding?

#6
by zoyer - opened

Using these params through llama.cpp it seems like the model always misses something. Sometimes a colon when creating objects or just missing completely quotes when writing strings. Anyone else?

~/llama.cpp/models/GLM-4.7-Flash-Q8_0.gguf --ctx-size 60072 -ngl 99 --temp 0.2 --top-k 50 --top-p 0.95 --min-p 0.01 --dry-multiplier 1.1

these are some mistakes it made:

const COLORS = {
    player: '#00fff5',
    ground: '#16213e',
    groundTop: '#0f3460',
    platform: #e94560, <--- NOTICE HERE
    coin: #fcdab7,
    hazard: '#ff2e63',
    bg: '#1a1a2e'
};
        } else if (enemies.length === 0) {
            // Wave Cleared
            wave++;
            gold += 50 + (wave * 10); Wave clear bonus <--- FORGOT TO ADD "//"
            updateUI();
            if (wave % 5 === 0) {
                // Bonus wave every 5
                enemiesToSpawn = 10;
                spawnInterval = 30;
            } else {
                startWave();
            }
        }

same issue here, I also tried with --dry-multiplier 0

yep tried as well. Seems somewhat broken, I'm guessing the non-quant model works well? I haven't tried it

Same issue here.

Unsloth AI org

Jan 21 UPDATE: llama.cpp has fixed a bug which caused the model to loop and produce poor outputs. We have reconverted and reuploaded the model so outputs should be much much better now.

You can now use Z.ai's recommended parameters and get great results:

  • For general use-case: --temp 1.0 --top-p 0.95
  • For tool-calling: --temp 0.7 --top-p 1.0

If you can test and let us know if you get better results? Thanks so much!

CC: @BuenosAir @mufeed @zoyer

nice work!

zoyer changed discussion status to closed

Sign up or log in to comment